Preparation of clinical cytological samples for subsequent analysis requires a sequence of steps that can be exceedingly time consuming if practiced manually. Society's emphasis on improving and economizing health care has created an increased demand on sample analysis to assist with diagnosis and to provide information concerning continuing treatment methodology. These demands require that the sample analysis be accurate and, in many instances, be performed quickly. Indeed, the modern clinical laboratory must be capable of preparing and analyzing samples quickly and accurately to assist with the diagnosis and treatment of a wide range of conditions and diseases.
A sample preparation framework requires various modules in order to process samples before further analysis can be performed. Generally, these modules can include conveyance, sample identification (accessioning) and subsequent tracking, vortexing, sample container preparation, specimen loading into a sample container, centrifugation, removing a centrifuged phase from the sample container, and forming an assay preparation for analysis.
Sample conveyance is the mechanism for transporting a sample between various modules. The means for sample conveyance have conventionally included both manual and automated transport systems. Automated systems can include conveyors or loading arms for facilitating the exchange of samples between processing stations.
Sample identification and tracking is particularly important to ensure custody transfer of cytological samples where the results will lead to a determination of the proper course of medical treatment for a subject. The potential for great physical harm exists if the samples become mixed or the results become improperly reported. Conventionally, samples are typically identified by a sample identifier.
Vortexing ensures the concentrations of components making up the sample are substantially continuous, cell agglomerations that may have formed are substantially broken up, and any entrained gas is substantially removed from the sample. Vortexing is particularly important when there has been a long delay between when the sample was taken and when the sample is prepared for further analysis.
Centrifugation is used to isolate particles in a suspended state from the medium in which they are held. Many research and clinical applications rely on the isolation of cells, subcellular organelles, and macromolecules typically from samples that need to be individually processed. Laboratory and/or clinical centrifugations conventionally are batch processes with most centrifuges designed to process multiple samples at once. When processing multiple, discrete samples, there is a delay in processing earlier samples placed into the centrifuge. This delay is known as dwell time. Furthermore, there must be a sufficient number of samples available to fill the centrifuge or at least there must be enough samples to load the centrifuge in such a way that the centrifuge remains in balance about its rotational axis once centrifugation begins.
Sample container preparation can involve any of a number of activities depending on, among other things, the type of centrifugation being performed and whether used sample tubes are discarded or recirculated.
Conventionally, a variety of means have been used to load a sample into a sample container. Sample loading can be accomplished by a sample transfer system that removes a sample from a container vial and dispenses the sample into a sample container. Sample loading can also include a manual or automated system and/or procedure for placing a sample container holding the sample into a desired position.
The supernatant or sedimentary layers of the centrifuged sample may be removed by a manual or automated system and/or procedure. Centrifuged sample portions may be unloaded by a sample transfer system that removes a sample portion from a sample container and dispenses the sample portion onto an assay device such as a slide or some other sample preparation used for analyzing the sample.
There remains a need in the art to more fully automate the sample preparation process. Eliminating the need for human intervention will allow samples to be processed more quickly; accurately; and, potentially, less expensively. A fully automated system will also reduce the amount of training that is required, reduce the size of the footprint of the automated preparation system, and increase the sample throughput per area of footprint. Further, an automated sample preparation system can reduce the amount of sample that is needed for processing. A fully automated system can also accommodate complete tracking of chain of custody from the sample vial through the final analysis. Further, there remains a need in the art to minimize human intervention reducing the potential that medical or laboratory personnel will come in contact with the specimen, contaminate the sample, or misdirect the sample through human error.
The extent of the idle time of a batch system capable of processing N samples but remaining idle until at least L samples are accumulated, with such samples arriving randomly to the batch system, has been addressed by Mathias A Dümmler and Alexander K. Schömig, “Using Discrete-Time Analysis in the Performance Evaluation of Manufacturing Systems” (paper presented at the annual International Conference on Semiconductor Manufacturing Operational Modeling and Simulation Meeting, San Francisco, Jan. 18-20, 1999). The amount of idle time is dependent upon both the number of samples, if any, remaining in the queue after the nth sequence starts and the number of samples arriving while the nth sequence is underway. The distribution of the number of samples remaining in the queue after the sequence has begun, yn(k), is given by:
            y              n        ⁢                                        ⁡          (      k      )        =      {                                                      0              ,                                                          k              <                              n                -                1                                                                                                                          ∑                                      i                    =                                          -                      ∞                                                                                                                                          ⁢                                          n                      -                      1                                                                      ⁢                                                                  ⁢                                  max                  ⁡                                      (                                          0                      ,                                                                                                    x                                                          n                              -                              1                                                                                ⁡                                                      (                            i                            )                                                                          -                        K                                                              )                                                              ,                                                          k              =                              n                -                1                                                                                                        max                ⁡                                  (                                      0                    ,                                                                                            x                                                      n                            -                            1                                                                          ⁡                                                  (                          k                          )                                                                    -                      K                                                        )                                            ,                                                          k              >                              n                -                1                                                        .      while the probability distribution of all prior samples waiting to be processed as they have accumulated at the end of the last sequence, xn-1 (k), is represented by:yn=max(0,xn-1−K)andK=L+min(max(0,xn-1−L),N−L).I.e., if the number of samples in the queue to be processed just prior to the nth sequence is greater than the number of samples that can be processed during the sequence, then these samples will wait to be processed in the next sequence. If there are an insufficient number of samples to either fill the batch system or meet the minimum number of samples required by the batch system before a sequence can begin, then there will be idle time in the operation of the batch system until a sufficient additional number of samples become available for processing.
Assuming geometrically distributed arrival times, the distribution of the number of samples arriving during any nth sequence, γn(k), is given by:
            γ      n        ⁡          (      k      )        =            ∑              m        =        k            ∞        ⁢                  ⁢                  (                                            m                                                          k                                      )            ⁢                                    p            k                    ⁡                      (                          1              -              p                        )                                    m          -          k                    ⁢                        b          n                ⁡                  (          m          )                    where bn(m) is the distribution along the length of the nth sequence and p is the probability of a sample arriving at any point in time.
The probability distribution of two random variables is given by the convolution theorem. Hence, the probability distribution for the number of samples waiting to be processed after the nth sequence, xn(k), is given by:
            x      n        ⁡          (      k      )        =                              y          n                ⁡                  (          k          )                    ⊗                        γ          n                ⁡                  (          k          )                      =                  ∑                  l          =                      -            ∞                          ∞            ⁢                          ⁢                                    y            n                    ⁡                      (            l            )                          ·                                            γ              n                        ⁡                          (                              k                -                l                            )                                .                    I.e., the number of samples waiting to be loaded after the nth sequence for the next n+1th sequence is dependent on the number of samples remaining in the queue to be processed, if any, just prior to starting the nth sequence and the number of samples that have arrived while the nth sequence is underway.
The mean time samples must wait before being processed, W, is given by Little's law:W=Q/Rwhere Q is the mean number of samples in the queue at the start of a sequence given by:Q=Σi·xn(i)and R is the average arrival rate of the samples.
Based on Little's formula, the mean waiting time of the samples before being processed, W, is minimized when there are consistently no samples waiting to be processed at the start of any sequence as long as there are at least a sufficient number of samples, L, available to be processed as required by the batch system.
The study provides revealing mathematical insight, using discrete time analysis, into the problems surrounding the potential limitations on batch processing in discrete time processing systems. As the analysis confirms, where the probability of appearance of a sample is reasonably consistent, then a batch system can be sized such that the idle time resulting from waiting for the requisite number of samples to arrive before a sequence starts can be minimized Indeed, where such probabilities are known, the batch system can be sized such that there are a sufficient number of samples to fill the batch system without any idle time between each sequence and any samples remaining at the end of a given period. However, such consistent probabilities in the clinical setting are rare. There will inevitably be variability in the probability of sample arrivals. Such variability typically is inconsistent and difficult to estimate. Hence, any batch system used in the clinical setting typically needs to be sized for those periods when the probability of arrival of samples is greatest in order to keep up with demand in those peak periods. Inevitably, this will lead to increased idle time when the probability of arrival of a sample is anything less than the maximum probability for which the batch system has been designed.
Conventionally, automated sample preparation systems provide the requisite modules in a subsystem form generally with each module functioning independent of the others within the system. Many of these modules operate in a batch-like manner further complicating the ability to streamline sample processing over the various processing sequences that occur within each of the modules. There remains a need in the art for an automated system that processes discrete cytological specimens wherein each module functions in an integrated manner.
Advancements have been made, for example, in the clinical laboratory to streamline sample processing and reduce the amount of sample that is needed on which to perform an analysis. The need to gain even further efficiency improvements from the sample preparation process has been recognized in the art. For example, U.S. Pat. No. 4,058,252 entitled “Automatic Sample Processing Apparatus” to Williams discloses advancing a number of centrifugation units each having a plurality of containers mounted on a conveyor to various processing stations. U.S. Pat. No. 6,060,022 entitled “Automated Sample Processing System Including Automatic Centrifuge Device” to Pang et al. discloses a centrifugation module that includes loading containers to be processed in a plurality of buckets, checking that the buckets are in balance, loading the buckets into the centrifuge, centrifuging, and unloading the buckets from the centrifuge. However, these systems are limited since the sample holders must be balanced before they are placed in the centrifuge—a process that can prove to be time consuming. Further, these systems are subject to idle time, depending on the availability of samples to be prepared, because the containers or buckets cannot be centrifuged until at least a minimum number of samples have been loaded in such a way that the centrifuge maintains balance. The extent of idle time in these batch processing systems can be determined by the discrete time analysis disclosed herein.
Automated loading and unloading procedures for samples by robotics are disclosed in, for example, U.S. Pat. No. 5,166,889 entitled “Robotic Liquid Sampling System” to Cloyd, U.S. Pat. No. 5,769,775 entitled “Automated Centrifuge for Automatically Receiving and Balancing Samples” to Quinlan, and U.S. Pat. No. 6,374,982 entitled “Robotics for Transporting Containers and Objects within an Automated Analytic Instrument and Service Tool for Servicing Robots” to Cohen et al. However, these automated processing techniques still require that some or all of the preliminary and subsequent sample processing steps be suspended or withheld until centrifugation is complete on the batch of samples being processed in the centrifuge. Furthermore, other modules operating in batch mode—such as the tube handling wheel described in the '889 patent, the weighing station and rack handling robot described in the '775 patent, or the sample handler module described in the '982 patent—can further affect the ability to quickly process samples and streamline the sample preparation operation.
While advancements have been made to streamline processing discrete samples in a cytological sample preparation system, there remains in the art a need to process a varying number of samples in a cytological sample preparation system while reducing, if not eliminating, the idle time of the system resulting from the batch processing of samples in the various modules of the unit.
An additional need that remains in the art is the ability to process irregular critical or emergency samples, otherwise known as STAT samples, that require priority handling without any substantial loss in efficiency of processing other discrete samples in the cytological sample preparation system.